Proc. Nati. Acad. Sci. USA Vol. 83, pp. 3042-3046, May 1986 Population Biology art: Variation in bower decorating style among male bowerbirds inornatus (cultural rnlsson/geographic variadon/behavior) JARED DIAMOND Physiology Department, University of California Medical School, Los Angeles, CA 90024 Contributed by Jared Diamond, December 30, 1985

ABSTRACT Courtship bowers of the bowerbird Ambly- BACKGROUND ornis inornatus, the most elaborately decorated structures Bowers are avenues, huts, or towers of sticks decorated with erected by an animal other than humans, vary geographically objects such as fruits, flowers, mushrooms, and stones. and individually. Bowers in the south Kumawa Mountains are Males often hold a decoration in their bill while displaying to tall towers of sticks glued together, resting on a circular mat of a female, and experiments confirm that bower structure and dead moss painted black, and decorated with dull objects such decorations influence females' choice of mate (7). Whereas as snall shells, acorns, and stones. Bowers in the Wandamen males are polygynous and contribute nothing to the female Mountains are low woven towers covered by a stick hut with an after copulation, females perform the whole effort ofbuilding entrance, resting on a green moss mat and decorated with a nest and rearing the young. Species comparisons show that colorful objects such as fruits, flowers, and butterfly wings. male bowerbirds with duller ornamental plumage build fan- Young males build simpler bowers, and adult males differ cier bowers (4, 8). Thus, during bowerbird the among themselves. Experiments with poker chips of seven female's attention has been transferred from ornaments ofthe colors offered as decorations showed that individual male's body to those ofhis bower. A. inornatus is the species prefer some colors over others, individuals and populations with the dullest plumage (males and females identical, nearly differ in these preferences, certain objects are placed in specific uniformly brown) but with the largest and most elaborately parts ofthe bower, and birds steal chips from neighbors. Bower decorated bower. style may be partly learned. Hence, geographically varying My studies in the Kumawa, Wandamen, and Fakfak bower styles may be a culturally transmitted trait, like human Mountains were performed in February-March 1981 and art styles. August-October 1983. Bower sites in these uninhabited mountains lay at elevations of1050-2075 m and were reached Male bowerbirds of and build deco- by cutting trails inland for 5-20 days from the coast. rated structures termed bowers to woo females. The most complex bower-in fact, the most elaborately decorated NATURAL STRUCTURE OF BOWERS structure built by an animal other than Homo sapiens-is that Geographic Variation. South Kumawa Mountains. Bowers of the Vogelkop Gardener bowerbird Amblyornis inornatus. were spaced several hundred meters apart on ridge crests at This species is confined to five remote mountains of Indo- sites well lit from the east, thus enhancing visibility during nesian New Guinea that are hard to reach for political and early-morning hours of peak display activity. Each bower logistical reasons: the Arfak, Tamrau, Wandamen, Kumawa, incorporated one to five saplings ("maypoles"), about each and Fakfak Mountains (1). Thus, only a few briefaccounts of ofwhich was erected a tower 2-3 m high ofinterwoven sticks bowers in two of these mountains (Arfak and Tamrau) have 20-90 cm long. Sticks were glued together at contact points been published (2-5). with a white sticky substance (saliva?). Surrounding the This paper describes bowers of the Wandamen population sapling(s) was an almost perfectly circular mat of black dead and of the recently discovered Kumawa and Fakfak popu- moss up to 15 cm thick, with a diameter of 1.3-2.1 m and with lations. Previous studies had shown that bower styles differ the coefficient of variation of diameter in a given bower only among the 14 bower-building species of bowerbirds (4, 6). It 3%. (How a only 25 cm long lays out such a perfect circle now turns out that styles also differ conspicuously between is unknown.) At the base of each sapling was a cone of the populations of A. inornatus. Within a population there are same moss used for the mat. smaller differences between the styles of adults and im- All bowers of adult males had six types of decorations on matures and of different adults. To confirm that these or near the mat and segregated by type: pandanus leaves, each up to 2 m long and weighing up to 60 g (the bird weighs differences arise from intrinsic differences among birds 120 g), dragged from a considerable distance and propped rather than merely local differences in objects available for against a maypole at an angle of 30-45° or else laid on the decorating bowers, I tested the decorating preferences of ground; black elytrae ofbeetles; separate piles ofhundreds of individual birds with standardized test objects: poker chips of dark brown acorns, grey or brown snail shells, and dark seven different colors. Birds' choice and placement of chips brown stones; and a pile of black sticks from tree ferns and agreed with many features oftheir use ofnatural decorations. a substory tree. A whole bower and its decorations weighed These bower style differences among populations beg the several tens ofkilograms. The piled sticks, elytrae, and moss question as to how an individual bird acquires the style mat and cone were naturally brown but in the bower were characteristic of its population. I suggest that bower style is painted glossy black, apparently with an oily black substance not wholly innate but is partly learned, hence constitutes a in the bird's excrement. [Bower painting has been described culturally transmitted trait. previously for six species with avenue-style bowers (4, 6) but not for species with maypole-style bowers.] The publication costs of this article were defrayed in part by page charge Central Kumawa Mountains. At a site 8 km by air north of payment. This article must therefore be hereby marked "advertisement" the bowers just described, and within the same mountain in accordance with 18 U.S.C. §1734 solely to indicate this fact. range, Franz Sadsuitubun found 25 bowers similar to the 3042 Downloaded by guest on September 24, 2021 Population Biology: Diamond Proc. Nati. Acad. Sci. USA 83 (1986) 3043

FIG. 1. (Upper) A Wandamen bower, showing hut, door, moss mat, red leaves and green fruit on the mat, and black bracket fungi on moss-covered root in foreground. (Lower) Bower owner on pile of black fungi facing door and hut, with red leaf, orange flowers, and red fruit outside hut, yellow fruits inside hut. (Photographs by J.D. and John Ratcliffe.)

south Kumawa bowers, except for being also decorated with described by local New Guineans familiar with them, they are colored fruits (red, yellow, green, blue, white, or black). similar in structure and decorations to the south Kumawa Fakfak Mountains. Bowers were not seen by me, but, as bowers and lack colored fruits. Downloaded by guest on September 24, 2021 3044 Population Biology: Diamond Proc. Natl. Acad. Sci. USA 83 (1986) Wandamen Mountains. Wandamen bowers (Fig. 1) dif- bowers, except for two acorns at one bower. Conversely, fered drastically from south Kumawa bowers. There were colored fruits, flowers, fungi, and leaves like those used in the only one or two towers only 0.4-0.8 m high, of sticks woven Wandamen Mountains were available within a few meters of but not glued together. Covering the towers was a woven south Kumawa bowers but were ignored. White karst stones stick hut 0.9-2.2 m in diameter with an opening ("door") that were widespread in the south Kumawa Mountains but were faced downhill regardless of the compass direction. The mat used at only one bower, whereas only two Wandamen birds and cones were of live green moss, not of dead moss painted used the widely available orange bracket fungi. Thus, deco- black. The sole decoration shared with south Kumawa rations are not a random sample ofavailable objects: different bowers was beetle elytrae. A large pile ofblack bracket fungi populations and individuals select on the basis of different was placed downhill from the door. Decorations of many rules. colors lay on the mat or inside the hut, but details differed among bowers (see below): orange or black fungi of several EXPERIMENTAL STUDIES OF BOWER types; red leaves; orange bark; red, orange, yellow, green, blue, grey, brown, or black fruits; red, orange, yellow, or DECORATION purple flowers; and red, yellow, green, or black butterfly As tools to identify decorating rules, natural objects have the wings. disadvantages of nonuniformity and hard-to-quantify avail- Multiples of the same object (e.g., red flowers) were ability. As standardized test objects of known availability, I loosely grouped, as were objects of similar color (e.g., orange used numbered, glossy, circular poker chips of hard plastic, fruits with orange fungi, red flowers with red leaves). Certain 4 cm in diameter, 7 g in weight, and of seven colors: red, decorations besides black bracket fungi tended to be at orange, yellow, blue, purple, lavender, and white. Bower- specific locations-e.g., red leaves, black fruits, and orange birds' preferences may depend not just on hue but also on flowers solely outside the door; butterflies solely and blue saturation, lightness, and surface characteristics. In addition, fruits mainly inside the hut. When I shifted the position of a their perceptions of color will depend on the spectrum of decoration, the bower owner either restored it to the original illumination, the background, and details of their visual position or else discarded it in the forest. Decorations system, which may differ from that of humans. Thus, though changed from day to day as birds replaced wilting flowers and I shall describe experimental results in terms of poker chip rotting fruits with fresh ones. hue and though the resulting preferences correspond well to Arfak and Tamrau Mountains. As described by previous many features ofnatural bower decorating, the interpretation authors (2, 5), bowers in these mountains agree with Wanda- may actually be more complicated. men rather than Kumawa bowers, except that Tamrau Weeding and Harvesting Tests. In the south Kumawa bowers use white decorations (sap and egg shells) as well as Mountains I placed chips on the mats of bowers SK1-SK5. colored ones, whereas Wandamen bowers lack white objects. Within 10-30 min, birds SK2-SK4 picked up all chips Age-Related Variation. Compared to bowers at higher regardless of color and discarded them in the forest ("weed- elevation, low-elevation bowers in the Kumawa and Wanda- ed" them), whereas SK1 and SK5 left all chips in place. I then men Mountains were smaller and simpler in structure and had placed three chips of each color near each bower, but no bird few and atypical or no decorations. For example, the lowest ever took chips to decorate the bower ("harvested" chips). south Kumawa bower (SKi) had the smallest mat and cone, These responses correlate with south Kumawa birds' failure had the lowest tower, and was unique in lacking decorations, to use natural colored objects as decorations, with SKi's and and its owner was sloppy about removing wind-blown leaves SK5's sloppiness at keeping the mat clean of fallen leaves, from the mat. The lowest Wandamen bower had no decora- and with SKi's presumed immaturity. tions, the second-lowest had only a few bracket fungi and no In the Wandamen Mountains I first tested the three hut, and the third-lowest had a small hut and spartan and lowest-elevation bowers, which were also the smallest, aberrant decorations (rotten fruit and two acorns). Among simplest, and least decorated and were inferred to belong to many New Guinea montane bird species (including two immatures. These three birds all failed to weed chips placed Amblyornis species) young males are known to occur at on their mats, regardless of color. W1 failed to harvest any lower elevations than adults (8-10), and young male chips placed nearby, and W3 did not begin to harvest until the bowerbirds are known to build rudimentary bowers com- fifth day of testing. pared to adults (11, 12). Hence, low-elevation bowers of A. The remaining Wandamen bowers were at higher eleva- inornatus probably belong to immature males. tion, had various colored natural decorations, and were Individual Variation. Bowers of presumably adult males at inferred to belong to adults. Weeding tests at bowers W4 and higher elevation varied in mat and hut diameter, tower height, W5 showed that both birds quickly expelled white chips number of maypoles, and ornaments, as noted previously for placed inside the hut; W4 also expelled yellow, while red, Tamrau bowers (4, 5). In the south Kumawa Mountains male orange, blue, and purple were left in place (lavender was not SK5 mixed grey and brown snail shells in the same pile, SK4 tested). In harvesting tests at bowers W4, W5, W8, and W9 placed them in the same pile but with the brown shells toward with sets of three chips of each color placed outside the hut the center of the bower, and SK2 and SK3 put the grey and at a distance of 1-5 m, all four birds repeatedly harvested brown shells in separate piles. Whereas almost all birds used almost all available blue chips, no white, and some of other beetle elytrae, Wandamen birds W9 and W10 made a separate colors to decorate the bower mat. The preference ranking pile of beetle heads, which other males discarded. W4 averaged over the four birds was blue > purple > orange > specialized in orange objects (fruit, flowers, seeds), W5 in red > lavender > yellow > white, but preferences for colors butterfly wings, W6 in yellow flowers, W8 in purple flowers, other than blue and white varied among individuals (Table 1). and W10 in orange fungus. The second preference was purple (W4, W9), purple nearly Relation of Decorations Chosen to Those Locally Available. tied with red (W8), or orange tied for first with blue (W5). The This geographic, age-related, and individual variation in percentage of available lavender chips harvested varied bower decorations is not just due to variation in locally among individuals from 0% to 75%; yellow, 0-66%; red, available objects. For instance, all birds had access to beetle 20-83%. heads along with elytrae, but most discarded the heads. When I placed chips near bowers W8 and W9 and remained Acorns and snails were commonly available in the Kumawa to watch, the owners promptly began to harvest colors in a and Wandamen Mountains and were used regularly by the temporal sequence matching approximately the final se- hundreds at south Kumawa bowers but never at Wandamen quence of proportions found harvested after 0.5-1 day. For Downloaded by guest on September 24, 2021 Population Biology: Diamond Proc. Natl. Acad. Sci. USA 83 (1986) 3045

FIG. 2. (Left) (Underexposed): bowerbird with blue poker chip in beak, selected from pile of chips of seven colors. (Right) Mat and door of bower decorated with red leaves, green and orange fruits (bottom left), blue fruits and red flowers (top), and poker chips. Note grouping of chips by color and position of blue chips inside hut. instance, W9 first harvested blue chips (Fig. 2 Left), then and red fruits and flowers from W6's bower, leaving only the purple, then orange, then lavender followed by red (or vice less preferred yellow flowers. versa), but never yellow or white. This temporal sequence is Interbower warfare was not confined to theft: when I virtually the same as W9's rank sequence of harvested trapped the adult male owner of a south Kumawa bower in a proportions (Table 1). net, immature males repeatedly raided the bower, pulled Theft and Destruction. At almost every bower some chips sticks from the tower, stole decorations, and messed up the that I set out during harvesting tests disappeared, suggesting mat. Since male mating success in bowerbirds depends on the the possibility of theft by another bower owner. Hence, I bower, the surest way for a male to impair a rival's sex appeal traced thefts by numbering chips. When I placed three chips is to wreck his bower, as also observed for other bowerbird of each color at bower W6, bird W5 stole within 3 hr all blue, species (11, 12). orange, and purple chips, two red chips, and no yellow, Wandamen birds also tried to steal other man-made objects lavender, or white; by the next morning, all yellow and two to decorate bowers, such as my yellow Kodak film cartons lavender; the remaining one lavender and three white were (placed next to yellow fruits), blue matchbox (placed next to not subsequently stolen. This preference sequence (blue, blue fruits), and black camera. Bird W9 hopped onto one of orange, purple > red > yellow > lavender > white) matches my field associate's shoes and tried to pull off his blue sock that of the and brown shoelace. harvesting test for bird W5 (Table 1). Bower W6 Positioning. Wandamen birds often grouped chips of the was unusual in that its sole conspicuous natural decorations same color with each other and with natural decorations of a were yellow flowers. The poker chip thefts suggest that bird similar color (Fig. 2 Right). In a few cases chips of different W5 may have been stealing the usual blue, orange, purple, colors were stacked, and the color preferred in harvesting tests was generally placed on top. Among harvested chips, Table 1. Color preferences of individual Wandamen bowerbirds colors that were preferred (as judged by the harvested Bird proportions in Table 1) tended to be placed inside the hut, with less preferred colors on the mat outside the door. For Color W1 W4 W5 W8 W9 instance, the percentages ofharvested chips placed inside the Blue 0 100 100 100 93 huts ofbirds WS, W8, and W9 were blue, 96% (Fig. 2 Right); Purple 0 45 79 87 63 orange, 79%o; purple, 77%; red, 68%; yellow, 48%; lavender, Orange 0 27 100 75 40 21%. The corresponding numbers for the proportions of Red 0 25 73 83 20 available chips harvested by these three birds are very Lavender 0 0 32 75 28 similar: 98%, 83%, 76%, 56%, 35%, and 33%, respectively. Yellow 0 0 66 20 0 Observations ofbirds showed that decorating decisions are White 0 0 0 0 0 not automatic but involve trials and "changes of mind." For Twenty-one poker chips (three each of seven colors) were placed instance, bird W9 initially brought red and lavender chips near a bower. Shown are the percentages of chips of each color inside his hut and then eventually placed the red chip next to "harvested" by each bird (= found on the bower's moss mat 4-21 a red fruit in front ofhis door but discarded the lavender chip hr later), based on 6-11 trials per bird. in the forest. W5 ignored available purple chips during his Downloaded by guest on September 24, 2021 3046 Population Biology: Diamond Proc. Natl. Acad. Sci. USA 83 (1986) first two tests, brought them inside the hut on tests 3 and 4, females might learn taste in bowers from older females (11, placed them outside the door on tests 5 and 6, and resumed 12). bringing them inside on tests 7-10. By this argument, geographically varying bower styles Comparison with Natural Decorations. These tests with might be a culturally transmitted trait, like human art styles poker chips show that A. inornatus males do select among and bird song dialects (13). For instance, young Kumawa potential decorations (possibly, but not provenly, by hue) and males would learn to gather snail shells and acorns, glue that different individuals and populations differ in their together stick towers, and paint the mat black, whereas selection criteria. Poker chips elicit decisions similar to those young Wandamen males would learn to gather fruits and elicited by potential natural decorations, as shown by nu- flowers, weave stick huts, and make a mat oflive green moss. merous parallels: the undiscriminating behavior of low- Individual style differences within a population may also be elevation (presumed immature) Wandamen and Kumawa partly learned: a young male that often watches at the bower birds; use of colored poker chips and natural objects by of an older male specializing in green butterflies may thereby Wandamen but not Kumawa birds; selective positioning learn to prefer green butterflies. These speculations can be (e.g., grouping by color or placing blue chips or fruits tested by studies ofcage-reared birds (what sort ofbowers do preferentially inside the hut); and selective theft. Since they build if they cannot learn by observation?) or by study man-made objects can be standardized and have experimen- of a wild population for many years (do new fads arise and tally determined availabilities, they may offer advantages for spread, as true for songs of birds and whales?). analyzing bowerbirds' decorating decisions. It is a pleasure to thank Ardy Irwanto, John Ratcliffe, and Franz IS BOWER STYLE A CULTURALLY Sadsuitubun for help with the fieldwork; Jack Hailman, Mary TRANSMITTED TRAIT? LeCroy, and Melinda and Stephen Pruett-Jones for valuable discus- sion; and the National Geographic Society and World Wildlife Fund How is geographic and individual variation in A. inornatus for support. bowers maintained? There are two linked but separate questions: how males learn in what style to build and how females learn what style to prefer in choosing a male. Against 1. Diamond, J. M. (1985) Emu 85, 65-91. 2. Beccari, 0. (1877) Ann. Mus. Civ. Genova, sl, 9, 382-400. a strictly genetic hypothesis (i.e., that bower style is wholly 3. Ripley, S. D. (1942) Trail ofthe Money Bird (Longmans Green, instinctive) stand the facts that populations of various moun- London). tains are almost identical morphologically despite their very 4. Gilliard, E. T. (1969) Birds ofParadise and Bowerbirds (Nat- different decorating styles (1) and that birds only 8 km apart ural History Press, Garden City, NY). in the Kumawa Mountains differ significantly in style (using 5. Gilliard, E. T. & LeCroy, M. (1970) Amer. Mus. Novit. 2420, vs. not using colored fruits). 1-28. I suggest that learning plays at least some role in bower 6. Cooper, W. T. & Forshaw, J. M. (1977) The Birds ofParadise style. Young male Satin and Macgregor's bowerbirds build and Bowerbirds (Collins, Sydney, Australia). bowers ineptly and with colors atypical ofadults, spend much 7. Borgia, G. (1985) Anim. Behav. 33, 266-271. 8. Diamond, J. M. (1982) Science 216, 431-434. time watching old males at bowers, and take 4-7 years to 9. Diamond, J. M. (1972) Avifauna of the Eastern Highlands of develop the typical adult bower style (11, 12). A role of New Guinea (Nuttall Ornithological Club, Cambridge, MA). learning forA. inornatus males is tentatively suggested by the 10. Pruett-Jones, S. G. & Pruett-Jones, M. (1986) NatI. Geogr. spartan, atypical bowers at low elevations, attributed to Res. 2, 87-105. young males, and by the repeated testing of colored chips 11. Vellenga, R. (1970) Aust. Bird Bander 8, 3-11. before settling on a final use. Young males may learn by 12. Pruett-Jones, M. & Pruett-Jones, S. G. (1983) Nat. Hist. 92, watching older males or by discovering through trial-and- 48-54. error what attracts females, or by both means. Females 13. Bonner, J. T. (1980) The Evolution of Culture in sometimes go about in groups to visit bowers, so that younger (Princeton Univ. Press, Princeton, NJ). Downloaded by guest on September 24, 2021